No Cover Image

E-Thesis 29 views 5 downloads

What works to protect disadvantaged children and their families: a linked routine data approach / AMRITA BANDYOPADHYAY

Swansea University Author: AMRITA BANDYOPADHYAY

  • 2026_Bandyopadhyay_A.final.71745.pdf

    PDF | E-Thesis – open access

    Copyright: the author, Amrita Bandyopadhyay, 2026. Copyright: the author, Amrita Bandyopadhyay, 2026

    Download (8.42MB)

DOI (Published version): 10.23889/SUThesis.71745

Abstract

Aim: This thesis conducts a data-driven, population-level investigation into risk factors of early-life vulnerabilities using linked routine administrative data, integrated and harmonised with health, education and socio-economic records.Method: The primary areas of vulnerability examined in this th...

Full description

Published: Swansea 2026
Institution: Swansea University
Degree level: Doctoral
Degree name: Ph.D
Supervisor: Kennedy, N.
URI: https://cronfa.swan.ac.uk/Record/cronfa71745
Abstract: Aim: This thesis conducts a data-driven, population-level investigation into risk factors of early-life vulnerabilities using linked routine administrative data, integrated and harmonised with health, education and socio-economic records.Method: The primary areas of vulnerability examined in this thesis include low birth weight, low school readiness, living in deprived areas, exposure to domestic abuse, early alcohol use, injury risk and mental health challenges. Data-driven models using advanced statistical methods (logistic regression, negative binomial regression, Cox hazard regression) and machine learning techniques (feature selection and decision trees) are employed to identify significant risk factors and their association with vulnerabilities. The Wales Electronic Cohort for Children Phase 4 has been established through this research, compiling health, education and social care data of children born or growing up in Wales.Results: Consistent risk factors for low birth weight, low school readiness or poor academic outcomes include children living in deprivation, and poor maternal mental and physical health. Lifestyle issues such as maternal smoking, clinically significant alcohol use and substance abuse within families further exacerbate these vulnerabilities.Results reveal that children at risk of adverse outcomes, including early alcohol use and domestic abuse exposure, have fewer routine primary care contacts and more frequent emergency healthcare interactions, indicating neglect and challenging family circumstances for these children.Conclusion: The findings demonstrate that data-driven methods can identify the signs of neglect and the associated vulnerable population from linked routine data early on in their life. This research has led to nine published papers, contributing to a strong evidence base for policies and practices aimed at improving the life chances of disadvantaged children and shaping their life trajectories.
Keywords: Early Years, Vulnerability, Data linkage, Data-driven, Machine Learning, Routine administrative data, Disadvantages children and families
College: Faculty of Medicine, Health and Life Sciences

Similar Items